@bobmatnyc/mcp-desktop-gateway

MCP Desktop Gateway

Universal MCP Gateway for Claude Desktop - Python-based bridge with built-in system automation tools and intelligent prompt training.

✨ Features

• 🚀 One-command installation via NPM
• 🔧 15+ built-in tools for system automation
• 🧠 Automatic Prompt Training - LangChain-powered continuous improvement
• 🎯 Smart Learning - Improves based on user feedback and errors
• 🌐 Cross-platform support (macOS, Linux, Windows)
• 🔐 Security features including command filtering and timeouts
• 📦 Zero configuration - works out of the box
• ⚡ Performance optimized with automatic bytecode compilation
• 🔌 Extensible architecture for custom connectors

🚀 Quick Start

Installation

# Install globally via NPM
npm install -g @bobmatnyc/mcp-desktop-gateway

# Auto-configure Claude Desktop
curl -sSL https://raw.githubusercontent.com/bobmatnyc/mcp-desktop-gateway/main/scripts/install-claude-config.sh | bash

# Restart Claude Desktop

That's it! The gateway is now available in Claude Desktop with all built-in tools.

Basic Usage

In Claude Desktop, you can now use commands like:

// Execute shell commands
execute_command(command="ls -la")

// Show system notifications (macOS)
system_notification(title="Hello", message="Task completed!")

// Get system information
get_system_info()

// List running applications (macOS)
get_running_apps()

🛠 Built-in Tools

Shell Tools (Cross-platform)

• execute_command - Run shell commands safely (script writing)
• list_directory - Browse the filesystem
• get_system_info - Get system information

AppleScript Tools (macOS only)

• run_applescript - Execute AppleScript code
• system_notification - Display system notifications
• get_running_apps - List running applications
• control_app - Control applications (activate, quit, hide)
• get_clipboard / set_clipboard - Clipboard management

Terminal Tools (macOS)

• terminal_execute_command - Execute commands with visual feedback
• terminal_new_tab - Open new Terminal tabs for parallel processes
• terminal_get_output - Monitor command output in real-time
• terminal_list_tabs - Manage multiple Terminal sessions

Prompt Training Tools

• rate_response - Rate the quality of responses (0.0-1.0)
• suggest_improvement - Suggest prompt improvements
• report_issue - Report problems with responses
• get_training_status - Check automatic training status
• trigger_training - Manually trigger prompt training
• get_training_history - View training history for prompts

Gateway Management

• list_connectors - Show active connectors
• gateway_health - Check gateway status
• hello_world - Test the connection

📋 Resources & Prompts

The gateway also provides:

• 15+ resources for accessing system information and training data
• 12+ prompts for guided assistance, workflow help, and training guidance

🧠 Prompt Training System

The gateway includes an advanced automatic prompt training system:

Key Features

• Automatic Feedback Collection from all user interactions
• Smart Training Triggers based on error rates, ratings, and feedback volume
• Four Training Approaches: Few-shot, Reinforcement, Meta-prompt, Adversarial
• Continuous Monitoring checks prompts hourly for improvement opportunities
• Safe Deployment with thorough evaluation and optional auto-deployment

Getting Started with Training

1. Set your OpenAI API key: export OPENAI_API_KEY=your-key
2. Initialize the system: python -m prompt_training.cli init
3. Training happens automatically based on usage patterns!

For detailed documentation, see Prompt Training README.

🔧 Configuration

Create a custom configuration at config/config.yaml:

server:
  name: "mcp-desktop-gateway"
  version: "1.0.0"
  log_level: "INFO"

connectors:
  - name: shell
    enabled: true
    config:
      timeout: 30
      working_directory: "/path/to/projects"

🧪 Development

Setup Development Environment

# Clone the repository
git clone https://github.com/mcp-desktop-gateway/mcp-desktop-gateway
cd mcp-desktop-gateway

# Set up development environment
make dev

# Run in development mode  
make run

# Run tests
make test

Development Commands

make help         # Show all commands
make run          # Run gateway in dev mode
make test         # Run test suite
make test-npm     # Test NPM package
make logs         # View logs
make clean        # Clean build artifacts

# Configuration management (3 simple commands)
make use-local-code  # Use MCP Gateway from local Python code
make use-npm-package # Use MCP Gateway from NPM package  
make use-original    # Restore your original config (eva-mcp-bridge, etc)

Version Management

# Show current version
python scripts/version.py show

# Bump version
python scripts/version.py bump patch    # 0.1.0 → 0.1.1
python scripts/version.py bump minor    # 0.1.0 → 0.2.0
python scripts/version.py bump major    # 0.1.0 → 1.0.0

# Create release tag
python scripts/version.py tag

🏗 Architecture

┌─────────────────┐         ┌──────────────────────┐         ┌─────────────────┐
│                 │   MCP   │                      │  HTTP   │                 │
│  Claude Desktop │ ◄─────► │    MCP Gateway       │ ◄─────► │ Custom          │
│                 │  stdio  │    (Python)          │ (opt)   │ Connectors      │
└─────────────────┘         └──────────────────────┘         └─────────────────┘

🔒 Security

• Command filtering prevents dangerous operations
• Configurable timeouts (max 60 seconds)
• Environment variable filtering removes sensitive data
• No sudo or administrative commands allowed
• Sandboxed execution environment

🐛 Troubleshooting

Common Issues

1. Python not found

   # macOS
   brew install python3
   
   # Ubuntu/Debian  
   sudo apt install python3 python3-pip python3-venv
   
   # Windows
   # Download from python.org

2. Permission errors

   sudo npm install -g @bobmatnyc/mcp-desktop-gateway

3. Gateway not connecting

   • Check logs: ~/Library/Logs/Claude/mcp-server-mcp-desktop-gateway.log
   • Run in debug mode: MCP_DEV_MODE=true mcp-desktop-gateway

📚 Documentation

• Project Documentation - Detailed architecture and API
• Development Guide - Development workflow
• Changelog - Version history
• Examples - Example connectors

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📈 Roadmap

• v0.2.0 - External connector HTTP API, connector marketplace
• v0.3.0 - Authentication system, rate limiting, metrics
• v1.0.0 - Stable API, production ready, enterprise features

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

Built for use with Claude Desktop and the Model Context Protocol.

📧 Support

• 📖 Documentation
• 🐛 Issue Tracker
• 💬 Discussions
• 📧 Email